Navigation auf uzh.ch
This pages lists the open BSc. and MSc. thesis descriptions, as well as the master projects opportunities currently available in the DDIS research group.
If you are interested in any of the listed projects, please do not hesitate to contact the person mentioned in the open topic description.
If there are currently no open topics but you are generally interested in our research (see https://www.ifi.uzh.ch/en/ddis/research.html), or if you would like to propose a thesis about your own idea, you can send us an email to firstname.lastname@example.org.
The huge amount of scientific literature makes it hard to keep track of relevant information and new hypotheses. Current approaches are limited in their ability to properly detect explicit hypotheses from articles, mostly due to the complexity of the language structures and the inadequate consideration of the context surrounding hypotheses. Assessing scientific literature and identifying hypotheses manually requires a significant amount of time and effort since researchers must read through many articles and examine them closely to locate the required information.
This master thesis aims to develop an active learning strategy for effectively detecting hypotheses in scientific articles. The study first assesses whether automatically detecting research directions facilitates hypothesis detection and leverages the context of research directions to create a targeted search for relevant hypotheses. In the second phase, the project adapts research direction classifiers to enhance hypothesis detection performance using active learning. This approach is expected to improve the likelihood of successfully extracting hypotheses from scientific literature.
Requirements: It would be ideal to have some basic knowledge of transformer-based models as well as Python and PyTorch programming skills.
Start date: ASAP
Contact: Rosni Vasu
Many voting advice applications such as smartvote.ch or voteview.com offer low-dimensional visualizations of the political landscape. Most often, voters or politicians are embedded in a 2D plane, such that individuals with similar political opinions are close to one another. These visualizations may vary depending on the choice of the dimensionality reduction algorithm.
This thesis will investigate the application of a specific dimensionality reduction approach that comes from political science: ideal-point estimation. It will use existing smartvote data from the 2019 Swiss National election. The goal is to investigate the complexity of different ideal-point estimation algorithms and compare their accuracy to well-understood machine learning tools.
Requirements: Basic data science programming skills in Python are necessary. Knowledge of R is helpful. It is advantageous to be familiar with dimensionality reduction algorithms like PCA or t-SNE.
Start date: ASAP
Contact: Fynn Bachmann